Colors given as tuple of ints are not working anymore for LinearColormap and StepColormap since v0.8.0

[Spyromain]

When I create a LinearColormap or a StepColormap with a list of tuples of ints as argument for the colors parameter, it is no longer converted to floats and leads to incorrect results.

Current behaviour (v0.8.0 v0.8.1):

>>> from branca.colormap import LinearColormap
>>> colormap = LinearColormap(colors=[(128, 64, 255), (255, 128, 0)])
>>> colormap.colors
[(128, 64, 255, 1.0), (255, 128, 0, 1.0)]
>>> colormap.rgb_bytes_tuple(0)
(32767, 16383, 65279)
>>> colormap.rgba_floats_tuple(1)
(255, 128, 0, 1.0)

Expected behaviour (v0.7.2):

>>> from branca.colormap import LinearColormap
>>> colormap = LinearColormap(colors=[(128, 64, 255), (255, 128, 0)])
>>> colormap.colors
[(0.5019607843137255, 0.25098039215686274, 1.0, 1.0), (1.0, 0.5019607843137255, 0.0, 1.0)]
>>> colormap.rgb_bytes_tuple(0)
(128, 64, 255)
>>> colormap.rgba_floats_tuple(1)
(1.0, 0.5019607843137255, 0.0, 1.0)

This seems to be due to the internal _parse_color function that was broken in 8a8a214#diff-6886114fc8e4fd6197705978c768e1f49469e2455fb1c4a8efda26a7b11bfa0bL51.

Current behaviour (v0.8.0 v0.8.1):

>>> from branca.colormap import _parse_color
>>> _parse_color((128, 64, 255))
(128, 64, 255, 1.0)

Expected behaviour (v0.7.2):

>>> from branca.colormap import _parse_color
>>> _parse_color((128, 64, 255))
(0.5019607843137255, 0.25098039215686274, 1.0, 1.0)

[Conengmo]

Sorry about that @Spyromain, and thanks for bringing this to attention!

[Conengmo]

@Spyromain I made a PR that should fix your issue. Would you be able to review it?

[Spyromain]

@Conengmo, thank you for fixing it!

Just wanted to point out that there is another bug that was already present in the previous versions and will still be there with this condition in the PR: https://github.com/python-visualization/branca/pull/192/files#diff-6886114fc8e4fd6197705978c768e1f49469e2455fb1c4a8efda26a7b11bfa0bR80.

Tuples of ints with only zeros or ones will not be normalized when converted to floats.

Current behaviour:

>>> from branca.colormap import _parse_color_as_numerical_sequence
>>> _parse_color_as_numerical_sequence((0, 0, 1))
(0.0, 0.0, 1.0, 1.0)

Expected behaviour:

>>> from branca.colormap import _parse_color_as_numerical_sequence
>>> _parse_color_as_numerical_sequence((0, 0, 1))
(0.0, 0.0, 0.00392156862745098, 1.0)

[Conengmo]

Great point! If a user provides ints between 0 and 1, we should assume they are bytes with a range of 0 - 255 and normalize accordingly. If they are floats between 0 and 1 we should assume they are already normalized. I'll update the code and add a test case for it.

[Conengmo]

@Spyromain I updated the PR to address the case you brought up. Could you perhaps take another look? Hopefully it's all good now, but if you see anything that's not as expected I'd love to hear it.

[Spyromain]

@Conengmo, seems all good to me, thanks again! 😀